CC BY
0Contents lists available at ScienceDirect
Journal of Cleaner Production
journal homepage: www.elsevier.com/locate/jclepro
Environmental impact assessment of regional switchgrass feedstock production comparing nitrogen input scenarios and legume-intercropping systems
Amanda J. Ashworth a *, Adam M. Taylor b, Daniel L. Reed c, Fred L. Allen a, Patrick D. Keyser d, Donald D. Tyler e
a Department of Plant Sciences, University of Tennessee, 2431 Joe Johnson Dr., 252 Ellington Plant Science Bldg., Knoxville, TN 37996, USA b Center for Renewable Carbon, Department of Forestry Wildlife and Fisheries, University of Tennessee, 2506 Jacob Dr., Knoxville, TN 37996, USA c Cumberland Habitat Conservation Plan, Department of Forestry Wildlife and Fisheries, University of Tennessee, 2506 Jacob Dr., Knoxville, TN 37996, USA d Center for Native Grasslands Management, Department of Forestry Wildlife and Fisheries, University of Tennessee, 2431 Joe Johnson Dr., 274 Ellington Plant Science Bldg., Knoxville, TN 37996, USA
e Department of Biosystems Engineering & Soil Science, University of Tennessee, 605 Airways Blvd., Jackson, TN 38301, USA
CrossMark
ARTICLE INFO
Article history:
Received 26 February 2014
Received in revised form
10 September 2014
Accepted 2 October 2014
Available online 14 October 2014
Keywords:
Switchgrass
Life cycle assessment
Nitrogen fertilizer
Legume inter-cropping
Biofuels
ABSTRACT
As the use of second-generation biofuel crops increases, so do questions about sustainability, particularly their potential to affect fossil energy consumption and greenhouse gas emissions. This study used a life-cycle approach to compare environmental impacts associated with three switchgrass (Panicum virgatum L.) production scenarios: i) regional production from a pool of Tennessee farmers based on in-field inputs and biomass yield; ii) varying nitrogen (N)-input levels from a replicated field study for 8-yrs i.e., a 100% and 9% decrease, and an 81% and 172% increase from 'baseline levels' of N inputs used under objective i; and, iii) a legume-intercrop system compared to baseline levels in order to determine effects of displacing synthetic-N with legumes. When compared across all agricultural inputs, nitrogen fertilizer production and breakdown resulted in the greatest environmental impacts. Although fertilization increased lignocellulosic yields, a 100% reduction in N-inputs from baseline levels reduced the formation of carbon, methane, and nitrous oxides per unit of production, (or dry tonne of biomass over 10-yrs) compared to a 172% increase. Switchgrass yield response indicated a 'less is more' scenario, as inputs beyond the current recommended input level (67 kg N ha-1) are not environmentally remunerating. During switchgrass biomass production, inputs with lesser impacts included phosphorus, herbicides, pesticides, and diesel fuel. Legume-intercropping reduced greenhouse gas emissions and groundwater acidification (5% and 27% reduction in global warming potential and formation of acidifying species, respectively) compared with the 67 kg N ha-1rate. Although N-fertilizers impact environmental sus-tainability of regional switchgrass feedstock production, environmental consequences can be reduced under proper N-management i.e., <67 kg N ha-1 or legume intercropping. However, given that the aim of second-generation feedstocks is to reduce the current reliance on fossil fuels, their production still requires fossil energy-based inputs. Consequently, greenhouse gas reductions and the extent of cleaner feedstock production during the agricultural biofuel supply chain is contingent upon input management and optimizing synthetic fertilizer usage.
© 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/3.0/).
Acronyms: LCA, life cycle assessment; GWP, global warming potential; GHG, greenhouse gas.
* Corresponding author. Tel.: +1 865 974 4962; fax: +1 865 974 4714. E-mail addresses: aashwor2@utk.edu (A.J. Ashworth), AdamTaylor@utk.edu (A.M. Taylor), danielreed@tennessee.edu (D.L. Reed), allenf@utk.edu (F.L. Allen), pkeyser@utk.edu (P.D. Keyser), dtyler@utk.edu (D.D. Tyler).
1. Introduction
There are growing concerns about the environmental sustainability of feedstocks, specifically input requirements and their relationship to the amount and value of the outputs. One tool for evaluating system environmental sustainability is life-cycle assessment (LCA), which measures inputs to and emissions from production life cycles (Cherubini and Jungmeier, 2010). In addition,
http://dx.doi.org/10.1016/j.jclepro.2014.10.002
0959-6526/© 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/3.0/).
as much as 21.3 million ha of existing agricultural land in the U.S. may be converted to perennial grass feedstocks (McLaughlin et al., 2002), thereby making regionally-specific environmental sustain-ability measurements, such as LCA, critical before large-scale adoption of a second-generation feedstock.
Life-cycle assessments include life cycle inventories and life cycle impact assessments (ISO, 2006). Life-cycle inventory (LCI) is an accounting of the energy and raw material inputs, as well as emissions to air, water, and soil (ISO, 2006). The life cycle impact assessment (LCIA) process characterizes and calculates the effects of emissions identified in the LCI into generalized impact categories. Impact categories at the midpoint level characterize impacts using indicators located along (but before the end of) the mechanism chain [e.g., parameter in a cause-effect network between the inventory data and the category endpoints (Bare et al., 2000; ISO, 2006)]. Examples of midpoint impact categories that are commonly used are global warming potential (GWP), photochemical ozone creation potential [POCP (i.e., smog)], acidification, eutrophication, and ozone depletion (NREL, 2011). Impacts of various production scenarios can be compared and thus LCA can be used to identify options that reduce a system's environmental impacts.
The majority of biofuel-focused LCAs have found a significant reduction in greenhouse gas (GHG) emissions from second-generation feedstocks, such as switchgrass, when compared with their conventional fossil fuel or first-generation counterparts [i.e., corn-(Zea mays L.)-ethanol] (Blottnitz and Curran, 2007; Cherubini and Jungmeier, 2010; Kim and Dale, 2005). The Brazilian paradigm for ethanol production [i.e. fermentation of sugarcane (Saccharurn officinarum L.)] is reportedly the highest performing firstgeneration biofuel crop in terms of environmental sustainability, and may exceed that of some second-generation feedstocks (Kendall and Yuan, 2013). However, composited, peer-reviewed LCAs determined that among all fuel pathways expected to contribute substantially to the U.S's renewable fuel portfolio, switchgrass-ethanol offered the greatest reduction in GHG emissions when compared to other fuel pathways [i.e., corn-ethanol, soy-(Glycine max L.) based biodiesel, waste/grease biodiesel, and sugarcane -ethanol] (Adler et al., 2007; EPA, 2009). In these analyses, switchgrass-ethanol resulted in a 124% and 128% reduction in GHG emissions (compared to a gasoline baseline) under two scenarios: 30 year, 0% Discount Rate (i.e., values all emission impacts equally, regardless of time of emission impact); and 100 year, 2% Discount Rate (discounts future emissions annually at 2%), respectively (EPA, 2009).
Net energy analyses have been used to evaluate fossil fuel and cellulosic biofuel-energy efficiencies. Reed et al. (2012) constructed an LCI of switchgrass fuel pellets based on surveys of switchgrass farmers and wood pellet producers in the southeastern U.S. and concluded that energy produced in switchgrass pellets was five times greater than the total fossil energy consumed to create them. Schmer et al. (2008) reported that switchgrass biomass yields ranging from 5.2 to 11.1 Mg ha-1 resulted in an average estimated net energy yield (NEY; or energy output ha-1 minus fossil energy input ha-1) of 60 GJ ha-1 yr-1 or 6.4 times more renewable energy than the non-renewable energy consumed. This study, as well as previous models [e.g., Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation Model (GREET) and the Energy and Resources Group (ERG) Biofuel Analysis Meta-Model (EBAMM)], assume a linear response of switchgrass yields to agricultural inputs. However, the majority of long-term field research indicates yield response reaches an asymptote before 134 kg nitrogen (N) ha-1, with an inflection point being reached at 67 kg ha-1 (Mooney et al., 2009). Other studies have reported negative energy balances for switchgrass-based ethanol under assumptions of high levels of
inputs (Pimentel et al., 2008); however, such assumptions are made without regionally measured plant-response values. Data are lacking that document specific environmental impacts from regional on-farm switchgrass production, especially impacts resulting from fertilization. Therefore, by evaluating impacts from varying fertilizer levels, the feasibility of environmentally sustainable feedstock production may be elucidated.
Production of second-generation biofuels may still result in acidification and ozone depletion due to use of synthetic fertilizers (compounds based on N and P) and pesticides (Larson, 2005; Zah et al., 2007). Nitrogen is an essential macronutrient in cropping systems but N emissions from its breakdown and production (e.g. nitrous oxide) are assumedly major contributors to GWP, acidification, and eutrophication in LCA simulations. Therefore, a positive result in one part of the cropping life cycle (i.e., increased crop yield) may be associated with a negative result (i.e. increased acidification) in another. However, there is a range of differences in production patterns and systems, as well as cultural practices, which makes regional and processing differences great (Ruviaro et al., 2012). Furthermore, the majority of LCAs conducted have been based on national or state-wide systems rather than on regional primary (enterprise-level) production data. Such regionally parameterized analyses of feedstock production are needed in order to more accurately model impacts of regional bioethanol production systems, albeit such specificity may preclude wide-scale assertions and recommendations.
A joint venture between the University of Tennessee, and DuPont Danisco Cellulosic Ethanol [i.e., University of Tennessee Biofuels Initiative (UTBI1)] established the nation's first pilot-scale, demonstration cellulosic biorefinery for the conversion of switchgrass to ethanol. This biorefinery, located in Vonore, Tennessee could require up to 154 tonnes of switchgrass per day and produce 19 million liters of ethanol per year (DOE, 2007). Since 2010, approximately 2100 ha of switchgrass have been planted, all of which are located within 80 km of the Vonore plant, making this the nation's first and largest switchgrass cellulosic ethanol venture. The study reported herein used: i) farm production-level field data from the UTBI project for modeling energetic efficiencies of switchgrass production, ii) an 8-yr, four sites field-study of four different N-fertilization rates on switchgrass biomass yield for a sensitivity analysis, and iii) a 4-yr replicated field study of intercropping legumes with switchgrass for biomass production as a replacement for the 'baseline N fertilization rate' used in objective i. Specifically, study objectives were to: i) identify the potential environmental impacts associated with farm-derived regional switchgrass production based on in-field agricultural input levels and biomass tonnage from a regional pool of farmers with the UTBI; ii) quantify environmental impacts as a function of decreasing or increasing N input levels. [i.e., a 100% and 9% decrease, and a 81% and 172% increase] compared to the baseline level of N input under objective i; and, iii) compare legume-intercrop systems to baseline results in order to determine the effects of displacing synthetic-N with intercropped legumes that host N-fixation.
2. Methods
This study applied LCA principles to evaluate environmental impacts of switchgrass grown under various scenarios. Cradle-to-gate LCI was developed (Reed, 2012) and LCIA used to compare production scenarios. This report is not however intended to present a complete LCA in full compliance with the ISO standards.
1 http://www.tennessee.edu/media/kits/biorefinery/docs/utbis_overview.pdf.
The agricultural production processes for both conventional and intercrop systems (Fig. 1, A1) include the basic flows and inputs in this analysis of farm-level switchgrass feedstock production. The 'baseline switchgrass production system' (A1) consists of the boundary assumed under objectives i and ii. The second production process is listed as 'legume-intercrop production system' in Fig. 1 (A1). The industrial conversion and processing system (A2) were outside the model boundary.
Assumptions for basic flow processes (depicted in Fig. 1) assumed that during harvesting, biomass is cut and field cured (<20% moisture). Drying beyond this point is considered an externality in this cradle to farm-gate LCA, as scenarios would likely not vary. Furthermore, this study did not consider storage facility or machinery construction. Soil carbon fluxes for all scenarios were also considered outside the system boundary. Finally, as fertilization increases both yields and impacts, the net impact per unit, or functional unit, allows for standardization. Therefore, this LCIA used 1 tonne (or 1 Mg) of dry weight switchgrass biomass as the functional unit (and is the standard unit in commerce).
Life cycle impact assessment was conducted using the Tool for the Reduction and Assessment of Chemical and other environmental Impacts (TRACI 2.0) model to determine midpoint indicators (Bare et al., 2003; USDA, 2013). TRACI is an LCIA program developed by the U.S. Environmental Protection Agency (EPA) specifically for the U.S. that uses input parameters for nine regions in the U.S. (Bare et al., 2003; EPA, 2002). All mid-point impact categories were included that were provided and supported by TRACI and SimaPro LCA software. Specifically, the environmental mid-point impact categories of global warming potential (kg CO2-eq), acidification potential (H+ moles-equivalent per kg of emission deposited), carcinogens (kg benzene-eq.), non-carcinogens (kg toluene-eq.), respiratory effects [particulate matter (PM 2.5-eq)],
eutrophication potential (kg N-eq.), ozone depletion [chlorofluo-rocarbons (kg CFC-11-eq.)], ecotoxicity (2, 4-D-eq.), POCP (kg NOx-eq.), and GWP (kg CO2-eq) were examined in this study. A range of environmental parameters were analyzed and described below.
Eutrophication characterization factors take into account the potential release of chemicals containing N or P into air or water per kilogram of chemical released, relative to 1 kg chemical discharged directly to surface freshwater (Bare et al., 2003). Acidification aims to predict impacts resulting from excess H+ ions in soil and water systems based on deposition in watersheds (Bare et al., 2003). The POCP midpoint category characterizes formation of ozone molecules in the troposphere and their resulting impacts on ecosystem and human health. An additional impact category reported is ozone depletion, which is the reduction of protective ozone in the stratosphere, generally caused by ozone reducing pollutants (e.g., CFCs, and halons); such reductions may result in greater ultraviolet radiation reaching earth's surface (Bare et al., 2003). In addition, GWP was examined or the potential change in earth's climate from a buildup of greenhouse gasses based on the gas heat retention capacity. In this study, the 100-year time horizon was used rather than GWP 500, which gives greater weight to longer-lived gases, as recommended by the Intergovernmental Panel on Climate Change (IPCC) and used by the U.S. EPA for policy-making.
2.1. Regional baseline switchgrass production scenario
Localized preliminary life cycle inventory (LCI) data were based on research from the UTBI (Reed, 2012). The LCI in-field data were derived from a survey sent to 61 participating producers (19% response rate or 12 respondents) in the southeastern region of Tennessee and provided input [e.g. diesel fuel (to operate machinery used during life cycle), seed, herbicide, fertilizer, etc.], and
Soil establishment: N fertilizer, K fertilizer, P fertilizer, lime, & diesel.
Switchgrass planting: N fertilizer, seedstock herbicide, pesticide,
Al: Agricultural production
Switchgrass harvest: electricity & diesel.
OUTPUT: 1 tonne of Switchgrass, (DW basis) Functional
Soil establishment: K fertilizer, P
Switchgrass planting: seedstock herbicide, pesticide, sufactant, &
Legume planting: seedstock & diesel.
Switchgrass harvest: electricity í diesel.
Cumulative System Boundary
Drying & Storage
Pretreatment & hydrolization
Sacchrification & co-fermentation
Combustion/co-generation
A2: Industrial conversion & processing
Fig. 1. Life cycle analysis system boundary, including both the agricultural production (A1) phases and the industrial conversion and processing (A2) steps.
switchgrass yield data for 2008, 2009, and 2010 (Table 1). Means from this survey established baseline yield and N-fertilization levels. Because some inputs (e.g., herbicide) decrease after perennial stand establishment, and yields increase, data were weight-averaged across surveyed farms to make up a ten-year life cycle period (assumed stand rotation). Switchgrass cv. 'Alamo' seeding rate (only for yr-1) was 7.8 kg pure live seed (PLS) ha-1. Per standard production protocols, no fertilization occurred during year one, as perennial bunchgrasses are establishing during this time and fertilization promotes weed growth and interspecific competition. Therefore, 2008 [stand establishment] data were weighted once, 2009 once and 2010 eight times because switchgrass yields reach an asymptote in year three, (attaining only 33-66% of its maximum yield potential in years one and two, respectively), with very little yield variation thereafter due to physiological maturity (McLaughlin et al., 2002; Mooney et al., 2009).
Farmer-derived agricultural input survey data (Table 1 ) for the inventory (LCI) collection was modeled with SimaPro LCA software (7th edition), which calculated the overall cradle-to-gate emissions associated with production processes (Fig. 1 ), by using a network of related inventories associated with regional inputs (Reed, 2012). Input data used in this study represent a 'moderate' scenario and are consistent with other energy consumption values for switchgrass biomass production (EUBIA, 2007; Pimentel and Patzek, 2005), and were collected in accordance with the Consortium for Research on Renewable Industrial Materials (CORRIM) research guidelines for life cycle inventories (CORRIM, 2001). Surveyed agricultural inputs included fertilizer e.g., N [urea (CH4N2O)], phosphorus [triple superphosphate (CaH4P2O8) (P)] and potassium [muriate of potash (KCl) (K)], diesel fuel, herbicide, surfactant, seed, and machinery for transportation and application of the aforementioned inputs. Other upstream processes in the reference system not explicitly examined in this research (e.g., equipment-use hours, electricity, diesel, and production of fertilizer, seed, herbicide, and surfactant, as well as GHG emissions emitted during these processes) were taken from the US LCI Database in SimaPro (Pré Consultants SimaPro, 2012; USDA, 2013). Producer yields were determined by biorefinery managers and University personnel, as switchgrass producers were paid on a per-tonne basis.
2.2. Switchgrass feedstock production under nitrogen input sensitivities
Because of the direct and indirect effects on the environment associated with fertilizing, sensitivity analyses were conducted to quantify the resulting environmental impacts at various input
Table 1
Cradle-to-(farm) gate life cycle inventory inputs per year per tonne of dry matter of switchgrass in the southeast. Input values are averaged over a 10-year production cycle (adopted from (Reed et al. (2012)).
Inputs Units Average valuea Coefficient of variationb
Diesel (tractor use) L 3.98 46%
Nitrogen (fertilizer) kg 4.77 84%
Phosphorous (fertilizer) kg 0.49 224%
2, 4-D (pesticide) L 0.05 219%
Glyphosate (herbicide) L 0.05 213%
Surfactant L 0.03 332%
Seedc kg 0.56 n/a
a Average value is the weighted average over a 10-year switchgrass stand rotation, where inputs and yields are assumed to be constant in years 3—10.
b Coefficient of variation is standard deviation/average of the reported data, without any weighting for stand age.
c Seed input values were not reported by farmers, however, a 7.84 kg ha-1 PLS rate was assumed once over a 10-yr life cycle.
levels based on a replicated, 8-yr small plot study on switchgrass N-input response data [e.g., a range of levels from low to a maximum deemed high enough to have surpassed switchgrass yield response; input values are considered typical nutrient response levels for this species (McLaughlin et al., 2002)]. Switchgrass cv. 'Alamo' nutrient response data used in the inventory analysis (objective ii) were collected by Mooney et al. (2009) at four different field sites at the University of Tennessee Research and Education Center at Milan (RECM; Table 2) for an 8-yr period (i.e. 2004—2011). The four sites were chosen to represent a range of soil types and landscape positions including two Grenada silt loam sites (well drained, level upland), a Vicksburg silt loam (a well-to moderately well-drained flood plain), and a Collins silt loam (a moderately-to somewhat poorly drained eroded sloping upland) (Mooney et al., 2009). Biomass yield data were collected from each site each year following the first-killing frost and weighted as described for objective i. In order to compile a 10-yr simulation period, yield data were averaged from 2006 to 2011 to make up the final 2 life-cycle years. No fertilizer was applied to any treatments at any site during yr-1 (e.g. establishment year). Starting in yr-2 the annual fertilizer treatments (yr-2 — yr-10) in this study were 0 (Sensitivity #1), 67 (Sensitivity #2; 60.5 kg N ha-1 yr-1 avg. over 10 yr; Table 2), 134 (Sensitivity #3; 121 kg N ha-1 yr-1 avg. over 10 yr), and 202 kg N ha-1 (Sensitivity #4; 181.5 kg N ha-1 yr-1 avg. over 10 yr) in the form of urea. From the results of this study, UT formulated the recommended rate of 67 kg N ha-1 yr-1 for switchgrass biomass production in Tennessee. Irrigation was not utilized in any field experiment and therefore, not included in any of the models, consequently data may not represent yield during years with extreme rainfall. Given the C4 photosynthetic pathway of this species, it is highly drought tolerant and therefore, irrigation is not needed for this crop in this region. In-field input data pooled and averaged from the 12 UTBI switchgrass farmer respondents (as described in Section 2.1, or 66.6 kg ha-1 averaged over a 10-yr production cycle; Table 2) served as the moderate-N or baseline scenario. Minimal yield fluctuation (±10%) occurred within each sensitivity model. Each sensitivity data set (8-yr yield data) was averaged per input level (Sensitivity #1—4). Life cycle simulations for the sensitivity analysis based on the aforementioned agricultural input/output data were run separately and compared to
Table 2
Measured annual switchgrass dry matter yield (±standard error) for sensitivity analyses (1—4) averaged across N-inputs for 10 yrs of production at University of Tennessee Research and Education Center (REC) locations (adopted from Mooney et al. (2009), at producer farms in conjunction with the University of Tennessee Biofuels Initiative (UTBI; Reed et al. (2012)) consisting of the baseline scenario, and REC centers for the legume intercropping scenario (Ashworth et al., 2012; Warwick, 2011).
Scenario N-input Relative to Yield Data
baseline source
—kgha-1- - --Mg ha-1- -
Sensitivitya# 1 0 -100% 5.68 1.9 REC
Sensitivity# 2 60.5 -9% 8.43 2.6 REC
Sensitivity# 3 121.0 +81% 9.18 2.9 REC
Sensitivity# 4 181.5 +172% 9.83 3.1 REC
Baseline level 66.6 0% 6.98 2.9 UTBI
Legume intercroppingb 0 -100% 6.98 2.9 REC
a Nitrogen input scenario sensitivities 1, 2, 3 and 4 (or 0, 67, 134, and 202 kgN ha-1, respectively, with 0 fertilizer in yr-1 for all scenarios) and baseline N-levels weight-averaged over 10 years of production (yr-1 received 0 kg ha-1 N for all sensitivity scenarios) and divided by yield data for 10-yrs in order for functional unit standardization.
b Legume intercropping scenario assumed: a 0 kg N ha-1 rate during the life cycle, yield was consistent with baseline results, seeding occurred every four years (2.5 times during a 10-yr life cycle) at 13.4 kg ha-1 PLS, and diesel inputs per planting period.
baseline level results for objective i. Sensitivity #2 (i.e. 67 kg N ha-1 per year after the establishment year or 60.5 kg N ha-1 avg. for the complete 10-yr life cycle) yielded 1.45 Mg ha-1 more dry matter compared to baseline results (66.6 kg N ha-1) likely due to robustness of data across climatic and phenotypic variation (i.e., 3 versus 8 data collection years). This default N-level used under objective i. was then compared to N-input sensitivities 1—4 i.e., a 100% and 9% decrease, and an 81% and 172% increase in order to monitor environmental impacts of agronomic inputs on a regional switchgrass biomass production basis.
2.3. Environmental assessment of switchgrass-legume intercropping scenario
Due to interest in N-input alternatives, a third scenario, or legume-intercropping was included. For the legume [i.e., red clover (Trifolium pratense)] intercrop system, several assumptions were made based on measured regional in-field data (Ashworth et al., 2012; Warwick, 2011). For example, based on local field studies by authors, every four years post-establishment, red clover requires re-seeding due to lack of persistence, and density being below proper levels (<4 plants per m2; Warwick, 2011) for targeted N fixation (67 kg N ha-1; biological nitrogen fixation determined via the 15-N enrichment method, data not shown). Consequently, for a 10-year simulation period, legume planting would occur two and a half times. Further, based on red clover-switchgrass intercropping studies, at proper legume densities switchgrass biomass yield does not vary from the 67 kg N ha-1 application (Ashworth et al., 2012). Consequently, the legume simulation assumed yield did not differ from the baseline scenario (objective i) when red clover was seeded at 13.4 kg ha-1 PLS. As such, input assumptions were as follows: legume seed and diesel for a single planting were multiplied by 2.5; a 0% yield reduction compared to LCA in objective i; a 13.4 kg ha-1 PLS legume seeding rate; same P inputs were required under objective i; and, 0 kg N ha-1 were needed over the entire simulation period (Ashworth et al., 2012). Other published clover inventory data with the SimaPro database (i.e. Ecoinvent v2.0; Jungbluth et al., 2007) were used for the upstream simulation of legume cropping systems under objective iii.
Fig. 2. Impact categories and their relative impact proportion per Mg of harvestable biomass based on system inputs for baseline production (UTBI switchgrass farmers, over a 10 yr simulation period).
(8000 kcal kg-1 N) (USDA, 2008). In this process, nitrogen gas is combined with hydrogen to form ammonia (NH3), and hydrogen in this reaction primarily comes from natural gas, one of the largest fuel inputs during this process. Current ammonia and urea production requires approximately 35 and 38 GJ per metric ton of N, respectively, in modern nitrogen plants (USDA, 2008). In addition, P and N inputs greatly influenced the eutrophication impact category (Fig. 2) by accelerating algal and aquatic weed growth and subsequent O2 limitations due to their decomposition, as P and N-limited aquatic systems respond strongly to minor nutrient increases. Such impacts from fertilization have been observed by others that have assessed environmental impacts from farmer-level inventories (Bojaca et al., 2014). Additionally, diesel inputs into switchgrass systems impacted POCP and ecotoxicity (approximately 50% of total impact per impact category) due to emission release to the air, water, and soil. Glyphosate production and utilization greatly impacted ozone depletion (i.e. 47% of total impact). Upstream processing and on farm utilization of both 2, 4-D and surfactant contributed less than 2.6 and 0.5% across all mid-point categories surveyed per unit of production, respectively. Production processes
3. Results and discussion
Figures illustrating the contributions to the total environmental impact per unit of production (Figs. 2 and 3) show the relative contribution rather than the importance of impacts [i.e. a small input can make a large contribution to a category that has a low total impact (such as grass seed input impacts for ozone depletion)]. The following figures show contribution analyses for the cradle-to-grave impacts of each system and have not been externally normalized (importance of impact category relative to external reference).
3.1. Environmental assessment of regional baseline switchgrass production
Greater than 50% of the total impact of inputs on regional agricultural switchgrass production inputs came from N and P fertilizer, diesel fuel, and glyphosate herbicide (Fig. 2). Of those inputs, nitrogen created the greatest deleterious impact on regional switchgrass production when compared across all impact categories, particularly respiratory effects, acidification, and global warming. This is largely due to the upstream manufacture of synthetic nitrogen fertilizer (i.e., Haber-Bosh) being energy-intensive, as breaking the trivalent bond of nitrogen (N=N) requires high pressure (100-200 atm), temperature (400-500 ° C), and energy
Fig. 3. Impact categories and their relative impact proportion for nitrogen input sensitivities 1, 2, 3 and 4 (or 0, 67, 134, and 201 kg N ha-1, respectively), legume intercrop, and baseline production scenarios from farmers with the University of Tennessee Biofuels Initiative (based on 10 yr simulation period and per Mg of harvestable biomass). A negative relative impact indicates a positive impact.
associated with grass seed inputs (e.g., largely upstream impact of diesel for harvesting, planting, and transport) resulted in ca. 20% of total ozone depletion relative to other inputs (Fig. 2).
3.2. Switchgrass feedstock production compared under nitrogen input sensitivities
Relative environmental impacts from N-input sensitivities compared to the UTBI Farmers (i.e. internally normalized to the baseline scenario) indicate multiple trade-offs on total cradle-to-farm gate impact categories (Fig. 3). In the figure, the various sensitivities were internally normalized (i.e. impacts for each sensitivity are expressed as a percentage of the highest-impact sensitivity).
Sensitivity #4 or the 181 kg N ha-1 rate (172% N increase relative to baseline level) resulted in the greatest (55% greater impacts relative to baseline scenario) acidification, carcinogenic, and eutrophication effects on a per-tonne basis, followed by Sensitivity #3 compared across all N-input levels relative to the baseline scenario (Table 2). Eutrophication reductions from the zero fertilization rate, compared to fertilized scenarios was also observed in previous studies (Borjesson and Tufvesson, 2011). As was expected based on the relative large impact of N fertilizer, Sensitivity analysis #1 (or a 100% decrease from baseline levels of 67 kg N ha-1) resulted in lowering the impacts across all categories relative to baseline except ecotoxicity, ozone depletion, and slightly positive impacts for smog production (+1.8 kg NOx-eq.) and non carcino-genics (+0.8 kg toluene-eq.); thus indicating N inputs result in greater yields (up to a point) but with disproportionate increases in environmental impacts. Consequently, inputs beyond the 67 kg N ha-1 level (Sensitivity #2), and perhaps even the 0 kg N ha-1 level (Sensitivity #1) may result in diminishing returns in terms of yield response and deleterious environmental impacts. Therefore, the current recommended rate of 67 kg N ha-1 is corroborated by environmental and agronomic efficiencies.
3.3. Environmental assessment of switchgrass-legume intercropping for nitrogen displacement
Results for the legume-intercropping life cycle assessment indicate a substantial portion of environmental disturbance from diesel fuel, phosphorus, and seed inputs on impact categories tested (Fig. 4). Namely, diesel inputs resulted in greater than 50% of the total impacts for acidification, non-carcinogenics, ecotoxicity, and smog; whereas under the baseline scenario, diesel was 20—51% less of a total constituent for the aforementioned mid-point categories, due to fossil-N contributing a greater portion of the total impact. Additional midpoint categories such as ozone depletion and GWP were affected by upstream seed inputs (approximately 60% and 42% of total impact, respectively). One likely explanation for seed input impacts were the inoculant coating and pre-treatment of legume seeds with water adsorbing polymers and limestone or rock phosphate that combat unfavorable soil and ambient conditions. Phosphorus inputs resulted in the greatest eutrophication constituent, likely due to P-limited algal systems and the resulting impacts of algal bloom proliferation under minor P increases, albeit phosphorus is not an overall large contribution in other impact categories and production systems at-large. Herbicide-related inputs (e.g., 2, 4-D, glyphosate, and surfactant) resulted in minor constituents [<14%) excluding glyphosate for ozone depletion (i.e. 35%)] of total impact for legume-intercropping systems (Fig. 4). Herbicide inputs for the legume intercropping simulation differed slightly compared to baseline results, as ozone depletion impacts from glyphosate decreased due to greater legume seed proportional impacts; whereas, impacts from
Fig. 4. Impact categories and their relative proportion of impact per Mg of harvestable biomass based on system inputs for legume-intercropping in switchgrass feedstock production systems in the Southeast U.S. (over a 10-yr simulation period). Seed inputs include both grass seed and legume seed.
glyphosate were less for ecotoxicity due to greater diesel fuel requirements for legume planting. Also, relative to farmers with the UTBI, legume-intercropping resulted in favorable impacts on acidification, carcinogens, and eutrophication, due to fewer N inputs and therefore less oxide of nitrogen emissions. However, non-carcinogens, respiratory effects, ozone depletion, ecotoxicity, and smog were greater under this system, likely because of greater diesel (upstream and downstream impacts) and legume seed (red clover) preparation and processing (mainly upstream impacts) during the 10-yr life cycle.
3.4. Climate forcing from regional baseline, nitrogen input, and switchgrass-legume intercropping scenarios
Global warming potential (kg CO2-eq) is among the most widely studied categories due to interest in climate change, renewable fuels, carbon credits and carbon sequestration. A 0 kg ha-1 of N (Sensitivity #1) rate over the 10-yr life cycle resulted in the least production of greenhouse gas emissions on a per-tonnage biomass basis over its life cycle compared across all analyzed scenarios (Fig. 5). Sensitivity #3 (or an 81% increase in N-inputs) resulted in
Fig. 5. Global warming potential (kg CO2-equivalents) of switchgrass production over a range of input levels (0, 67,134, and 202 kg N ha-1, Sensitivity #1—4, respectively), regional production from area growers, and a legume-intercropping scenario [internally normalized (divided by largest GWP score for all scenarios) and over a 10-yr simulation period and per Mg of harvestable biomass].
an 18% increase in GWP compared to UTBI, because yields were not greatly increased from the additional N (Fig. 5). Even though increasing fertilization generally increases crop production up to a point, the greatest N-rate input level (Sensitivity #4; or a 172% increase from baseline levels) had the highest formation of carbon, methane, and nitrous oxide on a per-tonne basis. Further, breakdown of synthetic-N in soils involves formation of N2O and consequently, this input level resulted in the greatest GWP (e.g. a 39 fold increase in N2O from sensitivity #1 to sensitivity #4). Secondly, the legume-intercropping simulation resulted in fewer stratosphere-warming gases, albeit only slightly less than the Sensitivity #2 (or the current recommended N-rate for Tennessee), because the legume-intercropping resulted in a substantial portion of CO2-equivalents from diesel fuel, P, and seed inputs. Further, diesel emissions (during planting and harvesting processes) and legume seed (including coating) was a major carbon-source for the legume-intercropping scenario compared to baseline results. Therefore, based on results, lower input levels resulted in a more carbon-neutral end-product on a per-tonne biomass basis.
Since nitrogen dioxide formation impacts the greenhouse effect and the formation of H+ ions, results from the GWP and aquatic acidification potential followed similar trends (data not shown). Formations of acidification species (i.e., sulfur and nitrogen oxides, hydrochloric and hydrofluoric acid, and ammonia) were greatest for the 202 kg N ha-1 rate. Conversely, the zero and current recommended nitrogen input rate had reduced acidification of groundwater potential on a per-tonnage basis. Furthermore, the legume-intercrop system caused less acid deposition than the UTBI simulation and the current recommended N rate (reductions of 54% and 27%, respectively).
4. Conclusions
The need for improved biofuel crop yields on a dwindling landmass due to population pressures has created conflicting solutions for environmental sustainability, considering nitrogen fertilization promotes enhanced biomass production, but is a carbon-positive input. Therefore the environmental sustainability of raw feedstock production is contingent upon agricultural production practices, and reducing the conventional reliance on fossil fuels in these systems can corroborate cleaner production. Environmental consequences still exist for switchgrass production, but could be augmented with management practices that sustain crop yields (e.g. sustainable nutrient management and legume intercropping), thus increasing greenhouse gas savings.
When compared across all inputs, the fossil-carbon-intensive process of producing nitrogen fertilizers resulted in the greatest deleterious impacts on all regional switchgrass production scenarios, resulting in greater respiratory effects, acidification, and global warming potential. This is due to the energy-intensive industrial N2 fixation process. Across all midpoint categories, there was a positive feedback when external inputs were reduced on a per-unit of biomass production basis. However there are still environmental consequences associated with low input (<67 kg N ha-1) switchgrass production systems, albeit likely still fewer consequences than first-generation feedstocks such as corn (which requires upwards of 168-250 kg N ha-1) for ethanol production (Kendall and Yuan, 2013), or fossil fuels such as diesel or gasoline (Bai et al., 2010). Consequently, there are potentially significant benefits offered by using switchgrass (with either the current recommended inorganic-N level, or legume intercropping nutrient regimes) as a bio-feedstock, particularly in terms of GHG emissions. However, these benefits may be offset by impacts in other categories such as respiratory effects, ecotoxicity, and photochemical creation, which have also been observed in other
assessments (Cherubini and Jungmeier, 2010). Consequently, management precision of switchgrass and other lignocellulosic crops requires consideration, and reduction of agricultural inputs could be pathways for advanced environmental impact reduction. Moreover, life cycle assessments of additional second-generation feedstocks are needed at the farm-scale to target biofuel crops with enhanced environmental performance in efforts to reduce the reliance on fossil fuel derived agricultural inputs.
Awareness of trade-offs, the diversity of cultural and processing practices, and sensitivities and assumptions are important in LCA models, and should be considered by policy makers. When compared across all N-input sensitivity scenarios, increasing fertilization in switchgrass feedstock production systems was less benign to human and environmental health on a per-tonne of feedstock production basis. Sensitivity #1 (no N fertilization) resulted in the lowest negative impacts across all categories, indicating that 'less is more' in terms of the environmental sustain-ability of switchgrass as a biofuel feedstock. However this was affected by switchgrass' minimal N-response beyond a modest threshold. Despite this result, Sensitivity #1 input level is not recommended due to the potential for long-term soil-nutrient depletion, likely resulting in reduced soil fertility and even yield loss over-time (>10 yrs or beyond the life cycle scope of this study). Based on inventoried field data, plant productivity began to reach a plateau (beyond Sensitivity #2) indicating increases in fertilizer inputs are not economically remunerating beyond the 67 kg N ha-1 nutrient input level for plant response (Mooney et al., 2009). This suggests a point of diminishing returns beyond this rate in terms of switchgrass N response and subsequent environmental impacts. Therefore, from an environmental and plant performance standpoint, sustainability of switchgrass feedstock production can be improved under proper N-management (i.e., <67 kg N ha-1; current recommended rate) or legume intercropping (lower potential GWP, carcinogenics, eutrophication, and acidification) at the farmgate level. However economic assessments of legume intercropping to determine breakeven points of this management practices are needed for switchgrass biofuel cropping systems to ascertain economic feasibility.
Data from this study can be included in regional databases for others interested in conducting LCAs, net energy ratios, or modeling feedstock systems. Producer surveys may allow for greater data representation and resolution of systems being modeled, however, there may be a tradeoff with lower sample sizes and lack of precision, as well as the inference space in which results may apply. Consequently, both wide-scale and localized environmental assessments of second-generation feedstocks are imperative going forward. Results herein provide regional guidelines for producing feedstocks sustainably, and may be used to identify potential improvements for production efficiency. Such data are needed to substantiate the claim that cellulosic-feedstock production systems are low-input and therefore environmentally sustainable.
Acknowledgments
Special thanks go to the USDA-NIFA Southeastern Region Sun Grant Program for funding the preliminary LCI on wood and switchgrass pellet production in the Southeast that was used in this project. We also extend our gratitude to the UTBI-contracted switchgrass farmers for their cooperation in providing vital infield data.
References
Adler, P.R., Del Grosso, S.J., Parton, W.J., 2007. Life-cycle assessment of net
greenhouse-gas flux for bioenergy cropping systems. Ecol. Appl. 17, 675-691.
Ashworth, A.J., Keyser, P., Allen, F., Bates, G., Harper, C., 2012. Intercropping Legumes with Native Warm-season Grasses for Livestock Forage Production in the Mid-South. University of Tennessee, Extension Bulletin. SP731-G.
Bai, Y., Luo, L., van der Voet, E., 2010. Life cycle assessment of switchgrass-derived ethanol as transport fuel. Int. J. Life Cycle Assess. 15, 468—477.
Bare, J.C., Hofstetter, P., Pennington, D., Udo de Haes, H.A., 2000. Midpoints versus endpoints: the sacrifices and benefits. Int. J. Life Cycle Assess. 5, 319—326.
Bare, J.C., Norris, G.A., Pennington, D.W., McKone, T., 2003. TRACI: the tool for the reduction and assessment of chemical and other environmental impacts. J Ind. Ecol. 6, 49—78. Massachusetts Institute of Technology, Yale University.
Blottnitz von, H., Curran, M.A., 2007. A review of assessments conducted on bio-ethanol as a transportation fuel from a net energy, greenhouse gas, and environmental life cycle perspective. J Clean. Prod. 15, 607—619.
Bojacä, C.R., Wyckhuys, K.A.G., Schrevens, E., 2014. Life cycle assessment of Colombian greenhouse tomato production based on farmer-level survey data. J. Clean. Prod. 69, 26—33.
Börjesson, P., Tufvesson, L.M., 2011. Agricultural crop-based biofuels — resource efficiency and environmental performance including direct land use changes. J. Clean. Prod. 19,108—120.
Cherubini, F., Jungmeier, G., 2010. LCA of a biorefinery concept producing bio-ethanol, bioenergy, and chemicals from switchgrass. Int. J. Life Cycle Assess. 15, 53—56.
Consortium for Research on Renewable Industrial Materials (CORRIM), 2001. Research Guidelines for Life Cycle Inventories. Available online at: http://www. corrim.org/pubs/reports/2010/phase1_interim/C0RRlMResProtocols.pdf (accessed and verified, June 2013).
United States Department of Energy(DOE), 2007. Switchgrass-to-Ethanol Plant Planned for Tennessee. Available at: http://apps1.eere.energy.gov/news/news_ detail.cfm/news_id=11323 (accessed and verified, October, 2012).
United States Environmental Protection Agency (EPA), 2002. Tool for the Reduction and Assessment of Chemical and Other Environmental lmpacts (TRACl). Available at: http://www.epa.gov/nrmrl/std/sab/traci/ (accessed and verified, January, 2013).
United States Environmental Protection Agency (EPA), 2009. EPA Lifecycle Analysis of Greenhouse Gas Emissions from Renewable Fuels. Available at: http://nepis. epa.gov/Exe/ZyPDF.cgi/P100B3F8.PDF?Dockey=P100B3F8.PDF (Accessed and verified, June, 2014).
European Biomass Industry Association (EUBIA), 2007. Available online at: http:// www.eubia.org/111.0.html. (accessed and verified, January, 2013).
lnternational Standards Organization (lSO), 2006. Environmental Management-life Cycle Assessment-principles and Framework. lSO 14040.
Jungbluth, N., Chudacoff, M., Dauriat, A., Dinkel, F., Doka, G., Faist Emmenegger, M., Gnansounou, E., Kljun, N., Schleiss, K., Spielmann, M., Stettler, C., Sutter, J., 2007. Life Cycle lnventories of Bioenergy. Ecoinvent report No. 17, v2.0. ESU-services, Uster, CH. from. www.ecoinvent.org.
Kendall, A., Yuan, J., 2013. Comparing life cycle assessments of different biofuel options. Curr. Opin. Chem. Biol. 17, 439—443.
Kim, S., Dale, B.E., 2005. Life cycle assessment of various cropping systems utilized for producing biofuels: bioethanol and biodiesel. Biomass Bioenerg. 29, 426-439.
Larson, E., 2005. A review of LCA studies on liquid biofuels for the transport sector. In: Scientific and Technical Advisory Panel of the Global Environment Facility (STAP) Workshop on Liquid Biofuels. New Delhi, India.
McLaughlin, S.B., De la Torre Ugarte, D.G., Garten Jr., C.T., Lynd, L.R., Sanderson, M.A., Tolbert, V.R., Wolf, D.D., 2002. High-value renewable energy from prairie grasses. Environ. Sci. Technol. 36, 2122-2129.
Mooney, D.F., Roberts, R.K., English, B.C., Tyler, D.D., Larson, J.A., 2009. Yield and breakeven price of 'Alamo' switchgrass for biofuels in Tennessee. Agron. J. 101, 1234-1242.
National Renewable Energy Laboratory (NREL), 2011. Life-cycle Inventory Database Project. Available at: http://www.nrel.gov/lci/ (accessed and verified, January, 2013).
Pimentel, D., Patzek, T.W., 2005. Ethanol production using corn, switchgrass, and wood; biodiesel production using soybean and sunflower. Nat. Resour. Res. 14, 65-76.
Pimentel, D., Williamson, S., Alexander, C.E., Gonzalez-Pagan, O., Kontak, C., Mulkey, S.E., 2008. Reducing energy inputs in the US food system. Hum. Ecol. 36, 459-471.
Reed, D.L., 2012. Life-cycle Assessment in Government Policy in the United States. PhD diss. University of Tennessee. Available at: http://trace.tennessee.edu/utk_ graddiss/1394 (accessed and verified February, 2013).
Ruviaro, C.F., Gianezini, M., Brandao, F.S., Winck, C.A., Dewes, H., 2012. Life cycle assessment in Brazilian agriculture facing worldwide trends. J. Clean. Prod. 28, 9-24.
Schmer, M.R., Vogel, K.P., Mitchell, R.B., Perrin, R.L., 2008. Net energy of cellulosic ethanol from switchgrass. Proc. Nat. Acad. Sci. 105, 464-469.
Pré Consultants SimaPro 7, 2012. Life-cycle Assessment Software Package, seventh ed. Amersfoort, The Netherlands. Available online at: http://www.pre.nl (accessed and verified, April, 2013).
United States Department of Agriculture, Foreign agriculture Service, 2008. Available at: http://www.fas.usda.gov/GATS/ExpressQuery1.aspx. (accessed and verified, January, 2014).
United States Department of Agriculture, 2013. Life-cycle Inventory Database Project. Washington, D.C. Available at: https://www.lcacommons.gov/nrel (accessed and verified, April, 2013).
Warwick, K., 2011. Establishment and Persistence of Legumes in Switchgrass Biomass and Forage/biomass Production Systems. MS Thesis. University of Tennessee. Available at: http://plantsciences.utk.edu/pdf/warwick_kara_thesis. pdf.
Zah, R., Boni, H., Gauch, M., Hischier, R., Lehmann, M., Wager, P., 2007. Life Cycle Assessment of Energy Products: Environmental Assessment of Biofuels. final report. EMPA—technology and society Lab, Auftrag des Bundesamtes für Energie, des Bundesamtes für Umwelt und des Bundesamtes für Landwirtschaft, Bern.
